Primos Storage Technology LLC v. Amazon.com Inc

I. Executive Summary and Procedural Information

• Parties & Counsel:
  • Plaintiff: Primos Storage Technology LLC (California)
  • Defendant: Amazon.com, Inc. and Amazon Web Services, Inc. (Delaware)
  • Plaintiff’s Counsel: Reichman Jorgensen Lehman & Feldberg LLP; Farnan LLP
• Case Identification: 1:25-cv-01571, D. Del., 12/30/2025
• Venue Allegations: Venue is alleged in the District of Delaware on the basis that both Defendants are incorporated in Delaware and have committed acts of patent infringement within the district.
• Core Dispute: Plaintiff alleges that Defendants’ cloud storage products and services, including Amazon S3 and Hadoop on Amazon EMR, infringe five patents related to transaction-based data storage systems and data coding methods for durability and recovery.
• Technical Context: The technologies at issue concern foundational methods for organizing, storing, and protecting massive datasets in large-scale, distributed computing environments, a critical component of modern cloud infrastructure.
• Key Procedural History: The complaint alleges that Defendants had pre-suit knowledge of the patents-in-suit, citing the prosecution history of an Amazon patent in which the U.S. Patent and Trademark Office cited Plaintiff's patents as prior art, and to which Amazon responded by distinguishing its invention.

Case Timeline

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 7,386,663 - “Transaction-Based Storage System and Method that Uses Variable Sized Objects to Store Data”

The Invention Explained

• Problem Addressed: The patent’s background describes the limitations of early 2000s file systems, particularly traditional "block-oriented" systems, which were slow, inefficient for large-scale distributed storage, and did not recover well from crashes Compl. ¶¶18, 20-21 These systems would overwrite data in fixed-size blocks, leading to scattered, slow write operations and system degradation over time Compl. ¶20
• The Patented Solution: The invention proposes a new type of file system that combines "transaction logging" with techniques from RAID storage systems Compl. ¶25 Instead of overwriting blocks, it writes variable-sized "objects" sequentially to a log that is striped across an array of disks Compl. ¶¶23, 27 This architecture is managed by a "rank manager" for configuring disk arrays, a "stripe space table" for tracking usage, and a "mapping component" to locate objects, all of which coordinate to log data transactions efficiently '663 Patent, abstract '663 Patent, col. 5:22-38 This approach is designed to increase write performance, enhance reliability, and allow for flexible expansion and crash recovery Compl. ¶27
• Technical Importance: The described solution offered a scalable, high-performance, and reliable file system architecture tailored to the demands of large-scale, distributed data storage that would become the foundation of cloud computing Compl. ¶27

Key Claims at a Glance

• The complaint asserts at least independent claim 1 Compl. ¶62
• The essential elements of independent claim 1 include:
  • A transaction-logging data storage system with a subsystem that stores variable-size data objects.
  • An array of data storage units configured into one or more "ranks," which provide "stripes" for storing the data objects.
  • A "rank manager" that performs configuration processing for the ranks.
  • A "stripe space table" that tracks space usage.
  • A "mapping component" that processes mapping information between object identifiers and their physical addresses.
  • At least one "processing component" that coordinates operations among the other components to implement the logging of data objects.
  • The mapping component includes subcomponents to enable the system to emulate common data storage abstractions (e.g., a key-value store).
• The complaint does not explicitly reserve the right to assert dependent claims for this patent.

U.S. Patent No. 7,937,528 - “Transaction-Based Storage System and Method that Uses Variable Sized Objects to Store Data”

The Invention Explained

• Problem Addressed: As with the related ’663 patent, this patent addresses the speed, reliability, and scalability limitations of conventional block-oriented and journaling file systems when applied to massive, distributed data sets Compl. ¶¶17-18, 20
• The Patented Solution: The patent describes a method of logging transactions in a data storage system that mirrors the system of the ’663 patent. The method involves storing variable-size data objects in a subsystem composed of striped arrays of disks ("ranks"), keeping track of space usage with a "stripe space table," processing mapping information with a "mapping component," and coordinating these operations to log both new data objects and changes to existing ones '528 Patent, abstract Compl. ¶104 This log-based approach avoids slow in-place overwrites and facilitates efficient crash recovery '663 Patent, col. 4:44-59
• Technical Importance: This method provided a procedural framework for a high-throughput, reliable, and scalable data storage system suited for the nascent cloud computing industry Compl. ¶27

Key Claims at a Glance

• The complaint asserts at least independent claim 1 Compl. ¶104
• The essential steps of independent claim 1 include:
  • Storing variable-size data objects in a data storage subsystem that uses an array of storage units configured into ranks and stripes.
  • A rank manager performing configuration processing.
  • Writing data corresponding to stripes into the ranks as a unit.
  • Keeping track of space usage via a stripe space table.
  • Processing mapping information between object identifiers and physical addresses via a mapping component.
  • Performing and coordinating operations among the subsystem, mapping component, and stripe space table to implement the logging of data objects and changes to stored data.
• The complaint does not explicitly reserve the right to assert dependent claims for this patent.

U.S. Patent No. 10,599,344 - “Transaction-Based Storage System and Method that Uses Variable Sized Objects to Store Data”

• Technology Synopsis: As part of the Transaction-Based Storage patent family, this patent addresses the same technical challenges as the ’663 and ’528 patents Compl. ¶¶16-27 Its claims focus on a system architecture that includes at least one "stripe buffer" for accumulating objects before they are written into a stripe, with objects being appended to the buffer according to a "write ordering policy" ’344 Patent, claim 1 Compl. ¶142
• Asserted Claims: At least claims 1 and 2 Compl. ¶142
• Accused Features: The complaint alleges that Amazon S3 utilizes an in-memory stripe buffer to accumulate data objects before writing them as a stripe across its "R2D2" server groups, and that objects are appended to this buffer based on factors like request timing and data durability requirements Compl. ¶148

U.S. Patent No. 8,078,944 - “Systems, Methods and Computer Program Products Including Features for Coding and/or Recovering Data”

• Technology Synopsis: This "Data Coding Patent" addresses the high cost and inefficiency of achieving data durability through simple replication in environments with frequent hard drive failures Compl. ¶¶30-32 The patented solution describes a method of "erasure coding" that transforms a set of "W" data inputs into a larger set of "H" output data streams using Galois field operations and a generator matrix. This allows the original "W" inputs to be recovered from any "W" of the "H" output streams, providing high durability with lower storage overhead than replication Compl. ¶34 The method includes producing an "H-sized intermediary" for each input to improve performance and applying check values for data integrity Compl. ¶¶35, 37, 188
• Asserted Claims: At least claims 1 and 2 Compl. ¶188
• Accused Features: The complaint alleges that Amazon S3 and Hadoop on Amazon EMR perform this method by using erasure coding to break data objects into "W" data shards and generate "H-W" parity shards, creating H total shards. The process is alleged to involve creating H-sized intermediaries and applying checksums to ensure data integrity Compl. ¶¶189-202

U.S. Patent No. 8,312,356 - “Systems, Methods and Computer Program Products Including Features for Coding and/or Recovering Data”

• Technology Synopsis: As the other "Data Coding Patent," this patent addresses the same data durability and efficiency problems as the ’944 patent Compl. ¶¶29-33 The claims are directed to a system comprising non-transitory computer-readable media containing instructions for a processor to perform the erasure coding transformation method, including processing "W" data inputs into "H" output data streams, creating H-sized intermediaries, and distributing the output streams between processing components Compl. ¶219
• Asserted Claims: At least claims 17 and 20 Compl. ¶219
• Accused Features: The complaint alleges that Amazon S3 and Hadoop on Amazon EMR are systems that operate on non-transitory media (e.g., memory) and execute software instructions that cause processors to perform the claimed erasure coding method, including creating and processing H-sized intermediaries and applying checksums Compl. ¶¶220-233

III. The Accused Instrumentality

Product Identification

• The complaint names two categories of accused products: "Transaction-Based Storage Patents Accused Products" and "Data Coding Patents Accused Products" Compl. ¶¶40, 54
  • Transaction-Based Storage Accused Products include Amazon Simple Storage Service (“Amazon S3”), Hadoop Distributed File System (“HDFS”) on Amazon Elastic MapReduce (“EMR”), computing instances running XFS (including on Amazon Elastic Compute Cloud (“EC2”) and Amazon Elastic Block Store (“EBS”)), and Lustre on Amazon FSx for Lustre Compl. ¶40
  • Data Coding Accused Products include Amazon S3 and HDFS on Amazon EMR Compl. ¶54

Functionality and Market Context

• The primary accused product described is Amazon S3, an object storage service marketed for its high durability, availability, and scalability Compl. ¶41 Its key technical features relevant to the allegations include "versioning," which maintains a persistent log of every version of an object, and "striping" or "sharding," where data objects are divided and distributed across arrays of storage devices to enhance performance and fault tolerance Compl. ¶¶42-43
• The complaint alleges that these technologies are critical to the success of AWS's cloud services and Amazon's broader business operations Compl. ¶2 The complaint also describes HDFS on EMR as a framework for distributed storage that uses a write-ahead log for transactions and employs striping for scalability Compl. ¶¶44, 46-47

IV. Analysis of Infringement Allegations

’663 Patent Infringement Allegations

• Identified Points of Contention:
  • Evidentiary Questions: The complaint's allegations regarding the specific functions of Amazon's internal, non-publicly named components (e.g., "C3PO," "Skynet," "R2D2") are made "on information and belief." A central point of contention may be whether discovery will produce evidence that these components function as the claimed "rank manager," "stripe space table," and "mapping component." The complaint supports these allegations with a detailed data flow diagram. This diagram shows the interaction between a user, a web server (WS), a blob assembler (BA), Skynet, C3PO, and R2D2s to process an image upload, illustrating the alleged roles of each component Compl. p. 29
  • Scope Questions: The analysis may turn on whether S3's "versioning" feature meets the claim limitation of a "transaction-logging data storage system." The patent distinguishes transaction logging from journaling file systems Compl. ¶¶22-23, and the court may need to construe the scope of this term based on the specification.

’528 Patent Infringement Allegations

• Identified Points of Contention:
  • Functional Equivalence: As with the ’663 patent, a key question will be whether the steps actually performed by the Amazon S3 service for object storage and versioning correspond to the specific sequence of steps required by claim 1, such as "keeping track of space usage via a stripe space table."
  • Scope Questions: The construction of "logging transactions" will be critical. The dispute may focus on whether creating a new version of an entire object, as S3's versioning does, constitutes the "logging" method described in the patent, which grew out of log-structured file systems.

V. Key Claim Terms for Construction

• For the ’663 and ’528 Patents:

  • The Term: "rank manager"

  • Context and Importance: This term is not a standard industry term and appears to be specific to the patent. Infringement hinges on whether Amazon's collection of internal management servers and processes (allegedly named "C3PO," "Skynet," etc.) meets the definition of a "rank manager." Practitioners may focus on this term because the complaint maps it to a combination of distinct Amazon components, raising questions about whether a single claimed element can read on multiple, distributed software processes.

  • Intrinsic Evidence for Interpretation:

    • Evidence for a Broader Interpretation: The claim language requires only that it "performs configuration processing associated with the one or more ranks" '663 Patent, claim 1 The abstract states a rank manager may be provided to "dynamically configure the ranks" '663 Patent, abstract This could support a construction covering any component or set of components that manages the configuration of disk groups.
    • Evidence for a Narrower Interpretation: The detailed description and figures illustrate specific functions, such as marking a rank as "read only," performing a "full rank evacuation" when a disk fails, and reconfiguring a rank without a failed disk '663 Patent, Fig. 7 A defendant may argue that a "rank manager" must be capable of performing these specific self-healing operations.

  • The Term: "stripe space table"

  • Context and Importance: The infringement allegation maps this claimed table to the function of Amazon's "C3PO servers" communicating with "R2D2 servers" to understand storage availability Compl. ¶68 The viability of this mapping depends on the construed scope of the term.

  • Intrinsic Evidence for Interpretation:

    • Evidence for a Broader Interpretation: The claim requires only that it "keeps track of space usage in the data storage subsystem" '663 Patent, claim 1 This could be construed broadly to cover any mechanism, including a distributed query process, that determines space availability.
    • Evidence for a Narrower Interpretation: Figure 11 of the patent depicts the "stripe space table" as a specific data structure that categorizes stripes into "Empty," "Almost Empty," and "Not Empty" based on a heuristic threshold '663 Patent, Fig. 11 A party could argue the term requires a distinct, persistent table with this specific categorical structure, not just a dynamic process for querying space.

VI. Other Allegations

Indirect Infringement

• The complaint alleges both induced and contributory infringement for all five patents. Inducement is primarily based on allegations that Amazon actively encourages infringement by marketing the accused products and providing documentation, user guides, and technical support that instruct customers on how to use the infringing features (e.g., S3 versioning, erasure coding in EMR) (Compl. ¶¶98-99, 136-137, 182-183, 213-214, 244-245). Contributory infringement is based on allegations that the accused products have no substantial non-infringing uses and are especially made or adapted for use in an infringing manner (Compl. ¶¶100, 138, 184, 215, 246).

Willful Infringement

• The complaint alleges willful infringement for all five patents. The primary basis for willfulness is alleged pre-suit knowledge stemming from the prosecution of Amazon's own U.S. Patent No. 9,772,787. The complaint asserts that during that prosecution, the Patent Examiner cited the publications of the ’663 and ’528 patents as prior art, and that Amazon studied the specification in detail to distinguish its invention, thereby demonstrating knowledge of the technology (Compl. ¶¶89-91, 127-129, 173-175). Similar allegations are made regarding prosecution of Amazon's U.S. Patent No. 8,935,221, where the ’944 patent was cited as prior art (Compl. ¶¶205-206, 236-237).

VII. Analyst’s Conclusion: Key Questions for the Case

• A core issue will be one of evidentiary mapping: can Plaintiff, through discovery, prove that Amazon’s internal and proprietary cloud architecture, which is described in the complaint based on "information and belief" using alleged codenames like "C3PO" and "R2D2," actually embodies the specific "rank manager," "stripe space table," and "mapping component" elements recited in the Transaction-Based Storage patents?
• A key question for the Data Coding patents will be one of operational equivalence: does Amazon’s implementation of erasure coding perform the specific transformation process claimed in the patents, particularly the step of creating "H-sized intermediaries," or does it achieve a similar result through a mathematically distinct method that falls outside the scope of the claims?
• A central legal question for willfulness will be the imputation of knowledge: can the act of distinguishing prior art cited by an examiner during the prosecution of one of a defendant’s own patents be sufficient to establish that the defendant "knew or should have known" that its commercial products risked infringing that same prior art, thereby supporting a claim of willful infringement?